Upload
brooke-dixon
View
215
Download
2
Tags:
Embed Size (px)
Citation preview
Computational Modeling of Anatomical and Functional Variability in Populations
Polina Golland
Computer Science and Artificial Intelligence LaboratoryMassachusetts Institute of Technology
Polina Golland, MIT CSAIL
Population Modeling
• Traditional Approach:– External information defines populations
• Images explain variability– Unimodal assumption: “average brain”
• Computational anatomy
• Our solution:– Images define populations
• External information correlates with image structure– Key idea: multiple templates
• Collaborators and Pubs:– R. Buckner (Harvard, HMS), M. Shenton (BWH, HMS)– Sabuncu et al. IEE TMI 2009.
Polina Golland, MIT CSAIL
Aging Study
• 400 subjects, ages 18-96– Some older subjects diagnosed with MCI
3 Templates:
Young OldMiddle
Polina Golland, MIT CSAIL
Age Distributions
2 Templates 3 Templates
Polina Golland, MIT CSAIL
Functional Geometry• Anatomy-free model of connectivity
– Use co-activation to embed in a functional space– Align embedded patterns across subjects
• Collaborators & Pubs:– A. Golby (BWH, HMS)– Langs et al. NIPS 2010, IPMI 2011.
Polina Golland, MIT CSAIL
Function Migration in Tumor Patients
Polina Golland, MIT CSAIL
• Unified model– Functional co-activations (fMRI)– Anatomical connectivity (DWI)– Population differences
• Collaborators & Pubs:– C.F. Westin, M. Kubicki (BWH, HMS)– Venkataraman et al. MICCAI 2010
Joint Model of Connectivity
Control Template
CA
CF
Schizophrenia Template
SA
SF
Polina Golland, MIT CSAIL
Connectivity Changes in Schizophrenia